v0.13.17

Fixes an issue with the resolver not mapping every reference that is a duplicate. It's an obscure bug, but was picked up by vacuum. No new features, just guaranteed reference resolving.

v0.13.16

Fix for stack overflow issue with resolving, when using a basepath and no actual files. No new features.

v0.13.15

Bugfixes to local file loader, resolver and mock renderer. No new features.

v0.13.14

Same as v0.13.13, but with an updated readme. Really just needed to kick sum.golang.org for being slow.

v0.13.13

Fixes the local and remote file system glitches and a regression added in v0.13.12

Also adds in some convenience methods for calculating size of the rolodex.

v0.13.12

Added the ability to locate where a reference comes from

• #177
• #200

A new method GetReferenceOrigin() exists on SchemaProxy that returns *index.NodeOrigin

type NodeOrigin struct {
  // Node is the node in question 
  Node *yaml.Node

  // Line is yhe original line of where the node was found in the original file
  Line int

  // Column is the original column of where the node was found in the original file 
  Column int

  // AbsoluteLocation is the absolute path to the reference was extracted from.
  // This can either be an absolute path to a file, or a URL.
  AbsoluteLocation string

  // Index is the index that contains the node that was located in.
  Index *SpecIndex
}

Also a new switch method features on the renderer.SchemaRenderer struct called DisableRequiredCheck() This will render all properties, regardless of the 'required` property values.

This release also tunes behavior of the local file system implementation used by the rolodex,

v0.13.11

A few bugfixes and tuneups for edge cases. No new features, no breaking changes

v0.13.10

No new features, small bugfixes for diff checking discovered by online users of openapi-changes

v0.13.9

No new features, just performance boosts and small bugfixes.

v0.13.7

The rolodex was indexing the entire basepath down (which is by design) by default. This is undesirable for some use-cases, because all kinds of non-project files are sucked in.

The new behavior indexes files recursively, and only reference files are inspected and indexed as they are discovered.

No breaking changes, no new features.